Introduction

The age of human-AI collaborative creativity is here. The writing in this zine except for this introduction was produced collaboratively with a state of the art language AI called GPT-3. The way it works is that you type in text into a box, and the AI will write something in response. In this zine the prompts are always in bold. The rest of the text is produced by the AI. This AI is easy to use and available to the general public, you can try it out at openai.com/api. GPT-3 was trained on 45 terabytes of text from across the internet. It is basically our culture reflected back at us. Imagine this scenario: we discover a lost city deep under the ocean. It has an extensive library, but the symbols are unknown to us. If we had an infinite amount of time to absorb that library, would we be able to write coherent text in that language? Would we be able to find out what it said? The AI is in this position, applying a generic pattern recognition algorithm to these terabytes of text, without any prior knowledge of language or the world. Astonishingly, it can produce coherent text. It not only uses correct syntax and semantics, but it can re-combine ideas in novel ways. We don't know if it "knows" what it's doing. When we interact with it, there are coherent words, which makes it very hard not to see it as some sort of communication. But who or what is communicating? Is the notion of communication an illusion? If an AI did want to communicate, how would it do so? If writing is not communication, then what is? What does it mean to be able to write without being human? We cannot answer these questions, but we have been fascinated, impressed, freaked out, puzzled, intrigued, and heartened by our interactions with the AI over these past months. We offer these writings as a document of our experiences. You can reach us at editors@gravenimages.ink. Erika Nesse